Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
1998-07-21
2001-08-14
Dorvil, Richemond (Department: 2641)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S201000
Reexamination Certificate
active
06275804
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention concerns a method for the storage of speech data in files in a digital dictation device with which the memory for the speech data is subdivided into clusters and each cluster is associated with a reference in a file allocation table to the next cluster of the file with the last cluster of the file being associated with a data end signal.
The grouping of stored data into differing units is known in the art of personal computers, in particular with the DOS operating system. The smallest information unit is a bit, with eight bits being grouped together into a byte and a particular number of bites (e.g. 512) forming a cluster. In this manner, data can be accessed not only bit by bit rather also in clusters. In particular for bulk storage media, this shortens the time required for changing the content of the memory (e.g. copying or deleting).
A contents register is also stored on the bulk storage medium in order to correlate a file to the physical storage location of the bulk storage medium. The contents register stores, for each file, the address of the cluster of the bulk storage medium first used for storage. In addition, a file allow cation table (abbreviated FAT) is stored on the bulk storage medium which contains the address of the subsequent cluster used to store the file in the address of the currently utilized cluster. The file allocation table FAT has a file-end entry in the last cluster used for storage, since no subsequent cluster exists.
Access to the memory region in which a file is stored is effected by searching the file name in the contents register of the bulk storage medium and the address of the first cluster stored in connection with the file name is utilized for access both to the file allocation table FAT as well as to the data in the file. Access to the FAT provides the address of the next cluster and the next entry in the file allocation table FAT. When the end of the file is reached, this entry contains a file-end mark. In this manner one recognizes that no further clusters belong to the file.
Files must also be stored in digital dictating machines containing speech recordings and their structure and management should be effected as in personal computers in accordance with the conventional DOS file formats in order to facilitate exchangeability of the storage medium between a portable dictating machine and a personal computer, however, with the modifications necessary for speech recording.
In speech recording using digital dictating machines one has, up to this point in time, been able to record the normally compress ed speech data without any problems including, at most, recording interruptions but otherwise always attaching new data to the current end of the file. This situation changes if so-called edit functions are to be facilitated. Such is the case with e.g. insertion of recordings at arbitrary locations into the existing recording as well as the deletion of arbitrary defined sections of an existing recording. These functions should of course be carried out in such a fashion that a subsequent play-back of the entire recording is effected with the individual sections being in the proper (logical) sequence and without interruptions.
In word processing systems this can be done by displacing the data of the existing record in correspondence with the insertion or deletion. Word processing systems can proceed in this manner, since text documents are usually completely loaded into the working memory of the personal computer where data manipulations can be rapidly facilitated. Since, however, speech data, in comparison to text data, require substantially more memory this process is very time-consuming and therefore user-unfriendly even in the event of a fast memory, particularly, however, with a slow bulk storage medium.
When storing speech data in a flash memory of a portable dictating machine, displacement of data is also extremely inefficient, since a flash memory must be erased prior to recording and the deletion, which can normally also only be simultaneously carried out for large memory regions, requires a relatively long amount of time and large amount of power.
It is therefore the purpose of the present invention to present a method which avoids the above mentioned disadvantages and which facilitates editing of data stored in accordance with the DOS standard in as rapid a manner and power-saving fashion as possible even if the files are stored on a relatively slow bulk storage medium. This should be guaranteed in particular when utilizing flash memories as bulk storage media.
SUMMARY OF THE INVENTION
This purpose is achieved in accordance with a method for the storage of speech data in files in a digital dictation device with which the memory for the speech data is subdivided into clusters, each cluster being associated with a reference in a file allocation table to the next cluster of the file. The last cluster of a file is associated with a data end-signal. The clusters are subdivided into speech frames and, in each cluster, validity information is stored for the speech frames stored therein, with the validity information indicating the storage location of the first speech frame of a cluster and how many speech frames are stored in the cluster. The remaining memory of the cluster stores encoded speech frames with the speech frames being marked with speech frame headers.
The dependent claims describe advantageous improvements.
In accordance with the invention, editing functions are carried out through manipulation of the references to the subsequent cluster to be output or to the subsequent speech frame to be output. Reference to a cluster having (logically) deleted data only is replaced by a reference to the subsequent cluster not containing deleted data. In this fashion, the chain of clusters is broken-up and reassembled in a manner shortened by the (logically) deleted data. The freed clusters are thereby either directly marked as being free or—in the case of a flash memory—indicated as “dirty” (released for deletion =potentially free).
In addition, for insertions, the necessary number of clusters can be easily added to clusters using the list linked together via the file allocation table in that, at the location where the insertion should be effected, the reference to the originally subsequent cluster is replaced with a reference to the first cluster which is to be inserted. For the last cluster to be inserted, reference is made to the originally subsequent cluster.
With speech frames which are to be deleted within a cluster, an additional reference to the first speech frame of the cluster and/or in addition the number of speech frames stored in the cluster is stored to thereby newly define the number of speech frames to be output. These references are stored at the beginning of each cluster at storage locations provided and reserved therefor. During output, one checks, in each case, for the last, therefore valid, entry for the address of the first speech frame and for the number of speech frames.
In order to guarantee most effective dictating including editing functions (insertion, gap-free deletion, recording-over dictations in a flash memory) in a cluster-oriented operating system using minimal amount of hardware (small working memory and limited processor capability), the following measures are, in particular, undertaken:
The cluster-oriented data system is adapted to the requirements through direct manipulation of the FAT and/or of one or more clusters. Cluster headers are utilized having entries for start addresses of the first speech frame and of the number of speech frames per cluster each having a plurality of entries for correction possibilities which are particularly useful in flash memory systems and having an entry for the speech encoding algorithm utilized.
This type of cluster header facilitates optimal memory usage by configuration of the speech frames in a cluster-independent fashion. In addition, a high memory resolution is achieved compared to the amount of memory required by t
Carl Holger
Kirchmann Dirk
Dorvil Richemond
Grundig AG
Vincent Paul
LandOfFree
Process and circuit arrangement for storing dictations in a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Process and circuit arrangement for storing dictations in a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Process and circuit arrangement for storing dictations in a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2503884